b'10 Partner News Spring 2023AcademicPredictive CAD System Suggestion for Efficient Engineering Design ReuseBy Gokula Vasantha, Jonathan Corney, John Quigley, and Andrew Sherlock, Edinburgh Napier.Predictive text systems complete wordsThe knowledge content of designs within by matching fragments against dictionaries.CAD databases has been extracted byDesignSimilarly, the vision of predictive CADidentifying common design structures (akaReuse(Computer Aided Design) is for systemsfeature snippets), which are a collection ofmeasurethat could complete models using shapefeatures (e.g., holes) with common parametricIntelligent Substitutablesearch technology to interactivelyvalues (e.g., diameters) that frequently occurSuggestion Featurematch CAD features against componenttogether. Furthermore, common designInterface Extractiondatabases. This way, an interactive,structures in connection with associationPredictivepredictive design interface would allowrules and geometric compatibility checksCADengineers to design new componentshave enabled the creation of an approach toSystemsmore effectively, incorporating establisheddiscovering substitutable features (2).Predictive Commonor standard, functional, and previouslyProbabilistic Designmanufactured geometries.An intuitive intelligent suggestion interfacemethods Structureformalizes 3D design activity as a markedDesignCommonalityThe goal of predictive CAD systems is topoint process that portrays the probabilitymeasureencourage engineers to actively reuseof specific features being added at designs, extract valuable knowledge fromparticular locations (3). The system learns CAD databases, and establish intuitivefrom a family of previous designs andMethods and Tools for Developing Predictive CAD systems (1)suggestion interfaces. The recentlygenerates inferences using a form of spatial completed collaborative Engineering andstatistics. Importantly, the system updates Physical Sciences Research Council-fundedthe probabilities every time a new feature project with the University of Strathclyde,is added, and the predictions become Edinburgh, and Edinburgh Napier hasmore accurate as a design develops. Also, demonstrated innovative approaches tothe study highlights that the adaptive these tasks that create the foundationsapplication of probabilistic predictive for developing successful predictive CADmodels in feature design reuse has systems. increased the suggestion precision.A Sample Common Design Structure of Hole Pattern in a Hydraulic Valve (2)To encourage design reuse, an automatedThe approaches mentioned above have reuse measure, based on the computationalestablished an innovative collection of mining of CAD data, is proposed thattools whose integration could underpin contrasts an efficient reuse process (i.e.,the development of future predictive CAD a higher concentration of use on fewersystems. Like predictive text systems, components) against a poor reuse scenariopredictive CAD systems will automatically (i.e. where each component family membergenerate queries and suggest appropriate is equally likely to be selected for use) (1).CAD features at any point while creatingSuggested CAD Feature Location hotspots based on The measure based on the KullbackLeibler3D models. Also, like predictive text, theProbabilistic based Marked Point Process (3)divergence helps to characterize the degreesystems algorithms can potentially learn of component reuse within a single product,from an engineers interactive selection of across a family of products, and at thethe suggestions proposed to improve the individual component family level effectively. quality of future suggestions.'